Optimization History¶
The Optimization History page visualizes the evolution of variables, objectives, constraints, and observables throughout the optimization process, showing how solutions improved over iterations.
Overview¶
This page displays time series plots showing: - How design variables changed during optimization - Evolution of objective values - Constraint satisfaction over time - Observable values throughout the process
Key Use: Understand optimization dynamics and convergence behavior
Features¶
Result Selection¶
Select Result: Choose which optimization result to visualize - Currently supports single result selection - All plots update when you change the selection
Note: This page requires results with iteration/time information
Variable Selection¶
Select which variables to display in each category:
Design Variables: - Input parameters that were optimized - Shows how the optimizer explored the design space - Default: First 3 variables selected
Objectives: - Optimization goals being minimized/maximized - Shows improvement over time - Default: First 3 objectives selected
Constraints: - Inequality and equality constraints - Shows feasibility evolution - Default: First 3 constraints selected
Observables: - Additional computed or measured values - Shows derived metrics over time - Default: First 3 observables selected
Plot Types¶
Time Series Plots¶
Standard Mode (for \<10,000 iterations): - Full resolution plots - All data points visible - Interactive hover and zoom
Resampled Mode (for >10,000 iterations): - Automatically activated for large datasets - Maintains visual fidelity while improving performance - Uses plotly-resampler for efficient rendering
Constraint Plots¶
Special visualization for constraints: - Feasibility Shading: Background shows feasible region - Threshold Lines: Shows constraint limits (0 for inequality) - Violation Highlighting: Clearly shows when constraints are violated
Understanding the Plots¶
Design Variables Plot¶
What it Shows: - How each design variable changed over iterations - Exploration patterns - Convergence behavior
Interpretation: - Flat lines: Variable converged early - Oscillations: Active search in that dimension - Trends: Directional improvement - Jumps: Major design changes or restarts
Objectives Plot¶
What it Shows: - Objective values at each iteration - Improvement trajectory - Convergence to optimal values
Interpretation: - Downward trend (minimization): Improvement - Upward trend (maximization): Improvement - Plateaus: Convergence or local optimum - Spikes: Exploration or constraint violations
Constraints Plot¶
What it Shows: - Constraint values over time - Feasibility status - Constraint violations
Interpretation: - Values ≤ 0: Feasible (for inequality constraints) - Values > 0: Constraint violation - Approaching 0: Active constraint - Far from 0: Inactive constraint
Visual Aids: - Green shading: Feasible region - Red shading: Infeasible region - Horizontal line at 0: Feasibility threshold
Observables Plot¶
What it Shows: - Evolution of computed or measured values - Derived metrics over time - System behavior indicators
Interpretation: - Depends on specific observable - Look for trends and patterns - Correlate with objectives and variables
Usage Workflows¶
Workflow 1: Convergence Analysis¶
- Select your result
- View objectives plot
- Identify when objectives stopped improving
- Check if optimization converged
- Assess solution quality
Look For: - Clear convergence to stable values - Continued improvement vs plateau - Oscillations indicating non-convergence
Workflow 2: Constraint Handling¶
- Select relevant constraints
- View constraints plot
- Identify when feasibility was achieved
- Check for persistent violations
- Understand constraint activity
Look For: - Transition from infeasible to feasible - Active constraints (values near 0) - Constraint violations over time
Workflow 3: Design Space Exploration¶
- Select design variables
- View variables plot
- Observe exploration patterns
- Identify important variables
- Understand optimizer behavior
Look For: - Variables that changed significantly - Variables that converged early - Exploration vs exploitation phases
Workflow 4: Multi-Objective Dynamics¶
- Select all objectives
- View objectives plot
- Observe trade-off evolution
- Identify when Pareto front was reached
- Understand objective conflicts
Look For: - Competing objectives (one improves, another worsens) - Synergistic objectives (both improve together) - Pareto front approximation over time
Best Practices¶
Variable Selection¶
Start with Objectives: - Always view objectives first - Understand overall performance - Identify optimization success
Add Variables Selectively: - Focus on most important variables - Too many lines create clutter - Add variables based on findings
Include Constraints: - Always check feasibility evolution - Understand constraint impact - Identify problematic constraints
Analysis Approach¶
Sequential Analysis: 1. Objectives: Did it improve? 2. Constraints: Is it feasible? 3. Variables: How did it get there? 4. Observables: What else changed?
Comparative Analysis: - Compare early vs late iterations - Identify phases of optimization - Understand algorithm behavior
Interpretation¶
Convergence Indicators: - Objectives plateau - Variables stabilize - Constraints satisfied - Observables steady
Problem Indicators: - Objectives not improving - Persistent constraint violations - Erratic variable behavior - Unexpected patterns
Common Patterns¶
Successful Optimization¶
Characteristics: - Objectives improve steadily - Variables converge to stable values - Constraints become satisfied - Clear convergence point
Example: - Initial exploration (high variation) - Improvement phase (objectives decrease) - Refinement phase (small adjustments) - Convergence (stable values)
Struggling Optimization¶
Characteristics: - Objectives oscillate without improvement - Variables jump around - Constraints repeatedly violated - No clear convergence
Possible Causes: - Poor algorithm settings - Difficult optimization landscape - Conflicting objectives - Tight constraints
Multi-Phase Optimization¶
Characteristics: - Distinct phases visible in plots - Sudden changes in behavior - Multiple convergence attempts
Interpretation: - Algorithm restarts - Phase transitions - Adaptive strategies
Advanced Features¶
Plotly Resampler¶
For large datasets (>10,000 iterations):
Benefits: - Maintains visual quality - Improves performance - Enables smooth interaction
Usage: - Automatically activated - Transparent to user - Zoom and pan work normally
Interactive Features¶
Zoom and Pan: - Click and drag to zoom - Pan to explore different time ranges - Double-click to reset
Hover Information: - Hover over lines for exact values - See iteration number - Compare multiple variables
Legend: - Click to show/hide variables - Double-click to isolate one variable - Useful for complex plots
Tips and Tricks¶
Visualization¶
Focus on Key Periods: - Zoom to early iterations for exploration phase - Zoom to late iterations for convergence - Compare different time ranges
Isolate Variables: - Use legend to hide/show specific variables - Focus on one or two at a time - Reduce visual clutter
Export Plots: - Download for reports - Include in presentations - Document optimization behavior
Analysis¶
Identify Phases: - Look for distinct behavioral changes - Mark important iterations - Understand algorithm strategy
Correlate Plots: - Compare variables with objectives - Link constraint violations to objective changes - Understand cause and effect
Validate Convergence: - Check if values truly stabilized - Look for continued small improvements - Assess if more iterations needed
Troubleshooting¶
Plots Not Showing¶
- Ensure result has iteration data
- Check that variables are selected
- Verify result is properly loaded
Too Many Lines¶
- Reduce number of selected variables
- Use legend to hide some lines
- Focus on most important variables
Can't See Patterns¶
- Try zooming to specific time ranges
- Adjust number of visible variables
- Check axis scales
Performance Issues¶
- Resampler should activate automatically for large datasets
- If slow, try reducing number of plots
- Close other browser tabs
Navigation¶
- Path:
/optimization-history - Category: Visualization
- Icon: Chart line icon
- Requires Data: Yes (with iteration information)
Related Pages¶
- Interactive Scatter: Explore variable relationships
- Parallel Coordinates: Multi-dimensional view
- Data Viewer: Examine specific iterations